You are on page 1of 23

This article was downloaded by: [201.141.53.

213]
On: 11 October 2012, At: 06:48
Publisher: Psychology Press
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office:
Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Cognitive Neuropsychology
Publication details, including instructions for authors and subscription
information:
http://www.tandfonline.com/loi/pcgn20

Inferring semantic organization from


refractory access dysphasia: Further
replication in the domains of geography and
proper nouns but not concrete and abstract
concepts
a
A. Cris Hamilton & Randi C. Martin
a
Psychology Department, Rice University, Houston, TX, USA

Version of record first published: 30 Nov 2011.

To cite this article: A. Cris Hamilton & Randi C. Martin (2010): Inferring semantic organization from
refractory access dysphasia: Further replication in the domains of geography and proper nouns but not
concrete and abstract concepts, Cognitive Neuropsychology, 27:8, 614-635

To link to this article: http://dx.doi.org/10.1080/02643294.2011.609541

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Any substantial
or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or
distribution in any form to anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation that the
contents will be complete or accurate or up to date. The accuracy of any instructions, formulae,
and drug doses should be independently verified with primary sources. The publisher shall not
be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or
howsoever caused arising directly or indirectly in connection with or arising out of the use of this
material.
COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8), 614– 635

Inferring semantic organization from refractory access


dysphasia: Further replication in the domains of
geography and proper nouns but not concrete and
abstract concepts

A. Cris Hamilton and Randi C. Martin


Psychology Department, Rice University, Houston, TX, USA

Patients with “refractory access dysphasia” have been a source of unique insight into the organization
Downloaded by [201.141.53.213] at 06:48 11 October 2012

of previously unexplored domains of semantic knowledge (i.e., proper nouns, geography, concrete and
abstract concepts). However, much of the relevant data have been based on the performance of a small
number of patients. Here, we present 2 patients who both display a “refractory access” pattern of per-
formance on spoken-word –written-word matching tasks and test their performance in the domains of
famous people, geography, and abstract and concrete words. While these patients show performance
similar to that for the previously reported patients in the domains of famous people and geography,
they show a very different pattern of performance with abstract and concrete nouns. We discuss
possible reasons why patients may differ in performance and evidence for and against the “differential
frameworks” hypothesis for the organization of concrete and abstract concepts.

Keywords: Semantic organization; Refractory access; Abstract concepts.

Patients identified as having “refractory access disparate words than when stimuli are grouped
dysphasia” are characterized by a unique pattern in arrays of semantically related words. It is this
of performance when confronted with matching- sensitivity to semantic relatedness that has been
to-sample tasks tapping language comprehension. exploited to make inferences about the organiz-
First, their accuracy decreases as a function of rep- ation of semantic knowledge. It is assumed that
etition. Second, their performance also declines items with greater semantic overlap (e.g.,
with faster presentation rates. For example, they leopard, lion, cheetah, puma) would generate
are worse with a 1-s response – stimulus interval greater semantic interference, while items with
(RSI) than with a 10-s or 30-s RSI. Finally, fewer overlapping semantic representations
these patients are remarkably sensitive to semantic would yield comparatively smaller interference
relatedness—their performance is much better effects. These effects have been observed in both
when stimuli are grouped with semantically spoken-word – written-word matching (Crutch &

Correspondence should be addressed to A. Cris Hamilton, Rice University, 474 Sewall Hall, Houston, TX 77005, USA (E-mail:
a.cris.hamilton@rice.edu)

614 # 2011 Psychology Press, an imprint of the Taylor & Francis Group, an Informa business
http://www.psypress.com/cogneuropsychology http://dx.doi.org/10.1080/02643294.2011.609541
REFRACTORY ACCESS DYSPHASIA

Warrington, 2005) and spoken-word – picture concrete and abstract words (Crutch &
matching (Forde & Humphreys, 1995, 1997; Warrington, 2005).
Warrington & Cipolotti, 1996). For example, Crutch and Warrington (2004)
It has been argued that patients with refractory reported patient A.Z. who was described as
access dysphasia differ in several fundamental ways gravely aphasic and demonstrated the three cardi-
when compared to patients with purely semantic nal features of a refractory semantic access deficit.
deficits (Jefferies, Baker, Doran, & Lambon, A.Z. was administered a spoken-word – written-
2007). Patients with refractory patterns are word matching task that presented names of
thought to have relatively spared semantic rep- cities that were organized by geographical proxi-
resentations, but mechanisms used to access mity (i.e., Brighton, Southampton, Eastbourne,
these representations are impaired. This is revealed Portsmouth) versus arrays that were geographi-
by the patients’ inconsistent performance with cally distant (i.e., Brighton, Norwich,
repeated presentation of stimuli. Patients may cor- Manchester, Plymouth). Patient A.Z.’s spoken-
rectly match spoken to written words on initial word – written-word matching was worse when
presentation, but then incorrectly match the arrays were composed of geographically proximate
Downloaded by [201.141.53.213] at 06:48 11 October 2012

same words when presented again only seconds cities than when they were geographically distant,
later. By comparison, patients with purely seman- presumably because the geographically related
tic deficits are reported to show much greater con- stimuli were more closely organized in semantic
sistency—they tend to miss the same items over space and shared more semantic representation
multiple trials and multiple testing sessions reflect- leading to greater semantic interference. These
ing the permanent loss of specific semantic rep- effects were recently replicated in two additional
resentations. Neuroanatomically, these patients patients by Crutch and Warrington (2010a).
differ in that refractory access patients typically A similar logic was used by Crutch and
have large lesions (often spanning frontal, parietal, Warrington (2004) who used a spoken-word –
and superior temporal areas), while patients with written-word matching task to test patient A.Z.’s
semantic deficits have typically sustained damage performance on famous people. Here, they used
to the anterior temporal areas following herpes famous names in arrays organized by occupation
simplex viral encephalitis (HSVE) or during the (i.e., politicians, broadcasters, etc.) compared to
course of the semantic dementia variant of fronto- arrays with famous names from different occu-
temporal dementia (FTD; Hodges, Patterson, pations. Patient A.Z. was significantly less accu-
Oxbury, & Funnell, 1992). rate when names were organized by occupation
Patients with semantic deficits following than when names belonged to different occu-
HSVE provided some of the first evidence that pations. Presumably, names organized more
categorical dissociations were possible following closely in semantic space yield greater interference
brain damage (Warrington & Shallice, 1984). effects, suggesting that occupation is one dimen-
However, recent data suggest that patients with sion by which famous people are organized in
refractory access dysphasia may allow an even the brain.
finer grained analysis of semantic categories. Another study with patient A.Z. (Crutch &
While patients with “category-specific” naming Warrington, 2005) examined the semantic organ-
disorders have typically demonstrated dis- ization of concrete and abstract words. As noted by
sociations between broad categories such as Crutch and Warrington, abstract concepts are
animals and artefacts or living and nonliving exceptionally difficult to study, given that they
things, patients with refractory access dysphasia are not straightforwardly depicted in pictures,
have been used to make inferences about the making them difficult to investigate using conven-
organization of semantic domains such as tional picture naming. When presented with a
famous people (Crutch & Warrington, 2004), spoken-word – written-word matching task,
geography (Crutch & Warrington, 2003), and Crutch and Warrington found that abstract

COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8) 615


HAMILTON AND MARTIN

words did not show the typical semantic related- However, the “differential frameworks proposal”
ness effects. That is, abstract words placed in offered by Crutch and Warrington (2005) is
synonymous arrays (i.e., divine, sacred, religious, unique in that it proposes different principles of
holy) did not elicit poorer performance than did organization for concrete and abstract concepts.
nonsynonymous arrays (i.e., divine, sparse, One might reasonably assume that differing prin-
furious, sudden). Abstract words are obviously dif- ciples of organization also suggests separate
ficult to assign to categories as they are not easily systems dedicated to concrete or abstract concepts,
grouped based on the dimensions that typically although Crutch and Warrington have contended
form categories for concrete words. However, that qualitatively different principles of organiz-
this null effect motivated Crutch and ation do not necessitate separate systems (Crutch
Warrington to explore other forms of related- & Warrington, 2010b). Interestingly, the pattern
ness—specifically, semantic association. Abstract of data reported by Crutch and Warrington
words were placed in semantically associated (2005) would seem to be at odds with numerous
arrays, in which items did not share a semantic cat- priming studies that have reported priming for
egory but instead were related based on “real- associatively related concrete items. This is
Downloaded by [201.141.53.213] at 06:48 11 October 2012

world or sentential contexts” (exercise, healthy, especially interesting given that both refractory
fitness, jogging). When compared to performance access effects and associatively based priming
on unrelated arrays, A.Z.’s performance on have been explained in terms of spreading acti-
abstract words presented in semantically associated vation (Collins & Loftus, 1975).
arrays was revealed to be significantly worse. Although data from refractory access dysphasia
Moreover, when concrete words were placed in patients have provided useful insights into the
associatively related arrays (i.e., farm, cow, organization of previously unexplored semantic
tractor, barn), no interference was observed rela- domains, much of this work has been done with
tive to unrelated arrays. Thus, Crutch and a very small number of patients. A subsequent
Warrington had established a double dissociation attempt to replicate the previously reported dis-
in patient A.Z.—her performance on abstract sociations between concrete and abstract data
words was influenced by semantic association, (Hamilton & Coslett, 2008) found a very different
while her performance on concrete words was pattern of performance, with refractory access
only influenced by semantic category. Crutch and patient UM-103 showing significant interference
Warrington argued that these data suggested for concrete and abstract words regardless of
that abstract and concrete concepts have funda- whether they were semantically related or associa-
mentally different principles of representation, tively related, suggesting the same principles of
with abstract concepts being organized associa- organization for both concrete and abstract con-
tively, while concrete words were organized by cepts. More recently, Crutch and Warrington
semantic category. However, they also noted (2010b) and Crutch and Jackson (2011) presented
that, “this associative/categorical dichotomy is additional patient data in support of the differen-
relative rather than absolute” (Crutch & tial frameworks hypothesis.
Warrington, 2005, p. 624). Clearly, there are contradictory data in the lit-
That concrete words and abstract words are erature, and it is thus valuable to examine
processed differently is not controversial. For additional patients on the concrete/abstract dis-
example, normal subjects are faster to make tinction, including the manipulation of semanti-
lexical decisions with concrete words (James, cally related versus associated items. In the
1975) and have better short-term recall for con- current study, we report data from two patients
crete words (Walker & Hulme, 1999). with aphasic profiles very similar to that of
Traditionally, theorists have proposed that abstract patient A.Z. That is, they are globally aphasic
words differ primarily in the number of semantic with moderately impaired comprehension and
features associated with them (Paivio, 1986). very little language production. Moreover, both

616 COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8)


REFRACTORY ACCESS DYSPHASIA

Figure 1. Patient H.A. Lesion was traced in native space, then transformed to MNI (Montreal Neurological Institute) space and displayed on
MNI template using MRIcron (Rorden, Karnath, & Bonilha 2007). To view a colour version of this figure, please see the online issue of the
Journal.

patients appear to possess the three principle fea- when provided a written word to match to the
tures of refractory access dysphasia. We then same pictures, scoring 97/100. A similar pattern
attempt to replicate the performance of patient was observed with stimuli from the
Downloaded by [201.141.53.213] at 06:48 11 October 2012

A.Z. in the domains of famous people and geogra- Psycholinguistic Assessments of Language
phy to establish that these cases are similar to pre- Processing in Aphasia (PALPA) version of the
viously presented cases. We then determine spoken-word – picture matching task (PALPA
whether these two patients show similar or differ- 47), where he was 32/40 (80%) and 39/40 (98%)
ing patterns of interference when presented with on the written-word – picture matching (PALPA
semantically related and associatively related con- 48). H.A. scored 94% on the picture version of
crete and abstract words. the Pyramids and Palm Trees task (Howard &
Patterson, 1992) and 88% correct on the written
Patient backgrounds version, a task tapping semantic processing (per-
formance at 90% and above being considered
Patient H.A.
normal). On an auditory lexical decision task
Patient H.A. was an 84-year old man who suffered
(PALPA 5), he was 88% correct, which was
a cerebrovascular accident (CVA) in 2006. H.A.
below the reported control mean of 95% correct
had earned a bachelor’s degree in college and
(Kay, Lesser, & Coltheart, 1992). He was 97%
worked in television broadcasting until his retire-
correct on a visual lexical decision task (PALPA
ment. His stroke resulted in a large left hemisphere
27), which was the same as the reported mean of
lesion that spanned frontal, parietal, and superior
controls. H.A. was very impaired at making two-
aspects of the temporal lobe (see Figure 1). H.A.
item visual homophone judgements (PALPA 15,
was globally aphasic with very limited output,
28/50 ¼ 56%), suggesting great difficulty in
such that naming tasks were not possible. For
accessing phonology from the printed word.
example, he failed to produce any scorable
response to the picture description of the
Western Aphasia Battery (WAB; Shewan & Patient D.Z.
Kertesz, 1980) and no correct responses to the Patient D.Z. was a 42-year-old woman who had
object naming subtest. H.A. was only able to completed a master’s degree and was enrolled in
repeat 4 of 14 words (29%) presented on the rep- a doctoral program when she suffered a CVA in
etition subtest of the WAB. His comprehension 2005. Her stroke resulted in a large left hemi-
of spoken words was also impaired. On the first sphere lesion that included frontal, parietal, and
100 items of the Peabody Picture Vocabulary temporal lobes (see Figure 2). D.Z. was also glob-
Test (PPVT; Form L; Dunn & Dunn, 1981), ally aphasic with extremely limited output. Much
which requires matching a spoken word to the like patient H.A. presented above, D.Z. could
correct item in an array of four pictures, H.A. not produce any output for the picture description
scored 68/100. However, he was much better subtest of the Western Aphasia Battery (Shewan

COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8) 617


HAMILTON AND MARTIN

Figure 2. Patient D.Z. Lesion was traced in native space, then transformed to MNI (Montreal Neurological Institute) space and displayed on
MNI template using MRIcron (Rorden, Karnath, & Bonilha 2007).

& Kertesz, 1980) and repeated only 3 of 14 words in Crutch & Warrington, 2005) and a 30-s RSI
(23%) presented on the repetition subtest of the (as reported in Forde & Humphreys, 1995, and
WAB. She scored 84/100 with spoken words on Warrington & McCarthy, 1983).
the first 100 items of the PPVT and scored simi-
Downloaded by [201.141.53.213] at 06:48 11 October 2012

larly on the written version (81/100). Similarly,


she scored 29/40 (73%) on the PALPA version Method
of spoken-word – picture matching (PALPA 47) Stimuli consisted of words sampled from four cat-
and 30/40 (75%) correct on the written-word – egories (fruit, vehicles, amphibians/reptiles, and
picture matching (PALPA 48). She was 83% musical instruments; see Appendix) and were pre-
correct on the auditory lexical decision task sented in both related and unrelated arrays. Arrays
(PALPA 5, controls ¼ 95%) and 90% correct on of four words were presented on a white sheet of
the visually presented lexical decision task 8.5 × 11′′ paper. The experimenter (A.C.H.)
(PALPA 27, controls ¼ 97%). On the Pyramids read a word aloud, and the patient was asked to
and Palm Trees task she showed impaired per- point to the word on the sheet of paper. Each
formance, scoring 80.7% correct with the picture word was probed once per cycle. Four pseudoran-
version and 84.6% correct with the written domized cycles were presented. Thus, each word
version. D.Z. was also very impaired at extracting was probed four times, creating 16 spoken-
phonology from printed words, scoring 22/50 word – written-word matching trials per array.
(44%) on visually presented homophone judge- Stimuli were presented with a 1-s, 10-s, and 30-s
ments (PALPA 15). RSI on three separate days. The experimenter
(A.C.H.) allowed 1 s between the patients’
responses and the presentation of the next word
EXPERIMENT 1 for the 1-s RSI and measured the 10-s and 30-s
RSIs with a hand-held timer. Administration of
Both patients were tested on spoken-word – the 1-s, 10-s, and 30-s RSI versions were separated
written-word matching task that presented words by several months.
in semantically related (all words from same
semantic category) or unrelated arrays. This exper-
Results
iment was intended to demonstrate that both
patients revealed the three essential features of a Data from Experiment 1 appear in Figure 3a.
“refractory access dysphasia”—that is, both Patient H.A. showed the typical effects of a refrac-
patients demonstrated worse performance with tory access dysphasia. He performed significantly
repeated presentation of the same stimuli, sensi- worse on the semantically related arrays (60.9%,
tivity to semantic relatedness, and better perform- 39/64) than on the unrelated arrays (89%, 57/
ance with increased RSI. With regard to RSI, we 64), x2(1) ¼ 13.5, p ¼ .0001, one-tailed. His per-
tested each patient with a 10-s RSI (as reported formance also declined with repetition, such that

618 COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8)


REFRACTORY ACCESS DYSPHASIA
Downloaded by [201.141.53.213] at 06:48 11 October 2012

Figure 3. Experiment 1. Patient H.A.: (a) Accuracy on related and unrelated trials collapsed over repetition cycles. (b) Errors for related and
unrelated trials at each repetition cycle. Patient D.Z.: (c) Accuracy on related and unrelated trials collapsed over repetition cycles. (d) Errors for
related and unrelated trials at each repetition cycle.

when results were collapsed over related and unre- ¼ 87.5%), x2(1) ¼ 6.564, p ¼ .005, one-tailed.
lated arrays, he was better at Cycle 1 (28/32 ¼ However, he showed no difference between the
87.5%) than at Cycle 4 (19/32 ¼ 59.4%), x2(1) 1-s and 10-s RSI (94/128 ¼ 73%), x2(1) ¼
¼ 6.488, p ¼ .005, one-tailed. Finally, H.A.’s 0.082, p ¼ .3875, one-tailed.
overall accuracy collapsed over related and unre- Patient D.Z. showed a similar pattern. She was
lated trials was worse with a short 1-s RSI (96/ worse with semantically related arrays (43/64 ¼
128 ¼ 75%) than with a long 30-s RSI (112/128 67.2%) than with unrelated arrays (60/64 ¼

COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8) 619


HAMILTON AND MARTIN

93.7%), x2(1) ¼ 14.366, p ¼ .0001, one-tailed. Results


Her accuracy also decreased with repetition. She
Both patients’ data appear in Figure 4. H.A. was
was 93.7% correct on Cycle 1 (30/32) and 75%
significantly worse on arrays composed of geo-
on Cycle 4 (24/32), x2(1) ¼ 4.267, p ¼ .0194,
graphically proximate cities (45/64 ¼ 70.3%)
one-tailed. D.Z.’s overall accuracy collapsed over
than on arrays consisting of geographically
related and unrelated trials was significantly
distant cities (58/64 ¼ 90.6%), x2(1) ¼ 8.401, p
worse with a short 1-s RSI (103/128 ¼ 80.4%)
¼ .0019, one-tailed. D.Z. showed the same
than with a long 30-s RSI (118/128 ¼ 92%),
pattern, being significantly worse on arrays with
x2(1) ¼ 7.447, p ¼ .0032, one-tailed. D.Z. failed
cities organized by geographical proximity (34/
to show any difference between the 1-s and 10-s
64 ¼ 53%) than on arrays of city names that
RSI (105/128 ¼ 82%), x2(1) ¼ 0.103, p ¼
were geographically more distant (50/64 ¼
.3744, one-tailed.
78.1%), x2(1) ¼ 8.87, p ¼ .0015, one-tailed.
Thus, both patients showed all three features of
Both patients H.A. and D.Z. showed similar
refractory access dysphasia—they were sensitive to
patterns to those for patient A.Z. reported by
semantic relatedness, showed an increase in errors
Downloaded by [201.141.53.213] at 06:48 11 October 2012

Crutch and Warrington (2003)—greater interfer-


with repetition, and had better overall accuracy
ence effects were observed when performing
with a 30-s RSI.
spoken-word – written-word matching with
arrays of cities that were geographically proximate
than with cities that were geographically distant.
EXPERIMENT 2: GEOGRAPHY
EXPERIMENT 3: FAMOUS PEOPLE
Crutch and Warrington (2005, 2010a) previously
proposed that place knowledge may be uniquely
A number of studies have suggested dissociations
organized by a spatial code. Support for this
between processing of proper nouns and
hypothesis came from experiments demonstrating
common nouns (see Semenza, 2009, for a recent
that city names presented in arrays composed of
review). For example, Cipolotti, McNeil, and
geographically proximate cities resulted in greater
Warrington (1993) reported a patient who
interference effects than did arrays composed of
showed better performance with famous people’s
geographically distant cities. Thus, we tested
names than with common objects. Moreover,
patients H.A. and D.Z. with similar stimuli to
some evidence suggests that famous proper
assess their correspondence with the previously
names might be a relatively preserved category of
reported patients.
knowledge in access dysphasia (Cipolotti &
Warrington, 1995). Finally, dissociations within
the category of proper nouns have been reported.
Method
For example, Forde and Humphreys (1995)
Patients H.A. and D.Z. were tested on city names reported a patient who was better with famous
using a spoken-word– written-word matching task names than with common first names (e.g., Bill,
using familiar city names sampled from different Don, Jack).
geographical regions. All cities were taken from Evidence that occupation may play an impor-
the continental United States, with four cities tant role in the semantic organization of famous
being sampled from the north-east (Boston, people comes from earlier studies examining
Philadelphia, Baltimore, New York), Midwest “covert recognition” of faces in prosopagnosia
(Chicago, Minneapolis, Detroit, Milwaukee), (De Haan, Young, & Newcombe, 1987). These
Texas (San Antonio, Dallas, Houston, Austin), studies required subjects to determine whether a
and California (Oakland, Los Angeles, San written name belonged to a specific occupation
Diego, San Francisco). (e.g., politician), but presented the names with

620 COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8)


REFRACTORY ACCESS DYSPHASIA
Downloaded by [201.141.53.213] at 06:48 11 October 2012

Figure 4. Experiment 2. Geographically proximate and distant cities. Patient H.A.: (a) Accuracy on geographically proximate and distant
arrays collapsed over repetition cycles. (b) Errors for proximate and distant arrays at each repetition cycle. Patient D.Z.: (c) Accuracy on
proximate and distant arrays collapsed over repetition cycles. (d) Errors for proximate and distant arrays at each repetition cycle.

pictures of faces sampled from the same occu- More recently, Crutch and Warrington (2004)
pation or different occupations. These prosopag- provided further evidence that occupation is one
nosic subjects showed interference such that they important dimension in the semantic organization
took longer to make decisions when names were of famous people by demonstrating that patient
presented with faces from different occupations A.Z. showed greater semantic interference effects
than when names were presented with faces in when written arrays of famous names were pre-
the same occupation. sented with famous names from the same

COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8) 621


HAMILTON AND MARTIN

occupation (i.e., Tony Blair, William Hague, experiments, each word was sampled four times.
Gordon Brown, John Major) than when arrays The next name was read after a 1-s delay.
were composed of names of famous people from
different occupations. Hamilton and Coslett
(2008) reported the same pattern with patient
Results
UM-103. Experiment 3 seeks to replicate this
effect with patients H.A. and D.Z. Data from both patients appear in Figure 5. When
initially presented with these stimuli, Patient H.A.
was worse with names that were arranged in arrays
Method
composed of names from the same occupation than
In Experiment 3, Patients H.A. and D.Z. were with items drawn from different occupations (same
tested on a spoken-word –written-word matching occupation ¼ 59%, different occupations ¼ 64%),
task that presented the names of famous people although this difference did not reach statistical sig-
from four different occupations. Given differences nificance. H.A. was tested two additional times sep-
in age between H.A. and D.Z., different sets of arated by several months, and when combining over
Downloaded by [201.141.53.213] at 06:48 11 October 2012

stimuli were constructed for each patient. H.A. all three administrations, he showed worse accuracy
was presented with baseball players (Babe Ruth, on items drawn from the same occupation (64%)
Micky Mantle, Sandy Koufax, Roger Maris), than from an unrelated occupation (72%), which
actors (Will Rogers, Charlton Heston, Burt was close to statistical significance, x2(1) ¼ 2.691,
Lancaster, and John Wayne), news broadcasters p ¼ .0504, one-tailed.
(Walter Cronkite, Tom Brokaw, Dan Rather, By comparison, Patient D.Z. was dramatically
and Ted Koppel), and U.S. presidents (Dwight worse with names from the same occupation
Eisenhower, Lyndon Johnson, Harry Truman, (accuracy ¼ 62.5%) than with names drawn from
John Kennedy).1 D.Z. was presented with different occupations (accuracy ¼ 95.3%), x2(1)
National Basketball Association (NBA) basketball ¼ 20.7, p ¼ .0001. Thus, both H.A. and D.Z.
players (Kobe Bryant, Allen Iverson, Michael demonstrated the pattern previously observed in
Jordan, Shaquille O’Neal), contemporary actors Patient A.Z. (Crutch & Warrington, 2004) and
(Tom Cruise, Brad Pitt, Russell Crowe, Bruce patient UM-103 (Hamilton & Coslett, 2008).
Willis), U.S. presidents (George Bush, Bill
Clinton, Ronald Reagan, John Kennedy), and
news broadcasters (Walter Cronkite, Tom
Brokaw, Dan Rather, and Ted Koppel). To EXPERIMENT 4
verify that the patients knew the occupations of
these famous names, we presented the mixed Crutch and Warrington (2005) argued that patient
arrays of famous people in a spoken-word – A.Z.’s performance on abstract and concrete words
written-word matching task, presenting the occu- suggested qualitatively different principles of
pation and having the patient point to the famous organization—abstract words organized by associ-
person that belonged to that occupation. Both ative relations and concrete words organized by
patients performed at ceiling (16/16 correct). In semantic category. Subsequent patient data both
the spoken-word – written-word matching task, support (Crutch, Ridha, & Warrington, 2006)
the experimenter (A.C.H.) spoke aloud each and failed to replicate this pattern (Hamilton &
name, and the patient was instructed to point to Coslett, 2008). Given the similarities between
the target from among four names printed on a patients H.A. and D.Z. to previously reported
page of 8.5 × 11′′ paper. As in previous cases (e.g., patient A.Z.; Crutch & Warrington,

1
H.A. indicated he had once been interested in baseball and film but had lost interest in contemporary fare; thus he was presented
with mid-twentieth-century baseball players and actors.

622 COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8)


REFRACTORY ACCESS DYSPHASIA
Downloaded by [201.141.53.213] at 06:48 11 October 2012

Figure 5. Experiment 3. Famous people from same occupation compared to different occupations. Patient H.A.: (a) Accuracy on related and
unrelated trials collapsed over repetition cycles. (b) Errors for related and unrelated trials at each repetition cycle. Patient D.Z.: (c) Accuracy on
related and unrelated trials collapsed over repetition cycles. (d) Errors for related and unrelated trials at each repetition cycle.

2003, 2005), Experiment 4 tested H.A. and D.Z. instructed to choose the corresponding word from
on the concrete and abstract stimuli. among four names printed on a page of 8.5 × 11′′
paper. The next item was presented after a 1-s delay.
Method
Results
Stimuli were adopted from the appendices of Crutch
and Warrington (2005). The experimenter (A.C.H.) Figures 6 and 7 present data from both patients in
spoke aloud each word, and the patient was Experiment 4. In contrast to data reported

COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8) 623


HAMILTON AND MARTIN
Downloaded by [201.141.53.213] at 06:48 11 October 2012

Figure 6. Experiment 4. Patient H.A.: Concrete and abstract words presented in semantically and associatively related arrays compared to
unrelated arrays. Column A: Accuracy for related and unrelated trials collapsed over repetition cycle. Column B: Errors for related and
unrelated trials at each repetition cycle.

624 COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8)


REFRACTORY ACCESS DYSPHASIA
Downloaded by [201.141.53.213] at 06:48 11 October 2012

Figure 7. Experiment 4. Patient D.Z.: Concrete and abstract words presented in semantically and associatively related arrays compared to
unrelated arrays. Column A: Accuracy for related and unrelated trials collapsed over repetition cycle. Column B: Errors for related and
unrelated trials at each repetition cycle.

COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8) 625


HAMILTON AND MARTIN

previously by Crutch and Warrington (2005) and the idea that both concrete and abstract words
Crutch et al. (2006), patients H.A. and D.Z. are connected by both semantically mediated and
showed a very different pattern of performance. associatively mediated relationships, but that con-
Both patients showed statistically significant inter- crete words depend more heavily on semantic
ference effects for concrete and abstract words relationships and abstract words more heavily on
when they were either associatively related or associative relationships. Neither patient demon-
semantically related. For the concrete words, strated any significant difference between concrete
H.A. was much worse with semantically related words in semantically related compared to associa-
arrays (61% correct) than with unrelated arrays tively related arrays—H.A., x2(1) ¼ 0.016, p ¼
(90% correct), x2(1) ¼ 28.823, p ¼ .0001, two- .449, one-tailed; D.Z., x2(1) ¼ 0.021, p ¼ .443,
tailed2. For the associatively related concrete one-tailed—or abstract words in associatively
arrays, H.A. again showed interference, with related compared to semantically related arrays—
worse performance on the related (59%) than on H.A., x2(1) ¼ 0.484, p ¼ .243, one-tailed; D.Z.,
the unrelated condition (77%), x2(1) ¼ 9.554, p x2(1) ¼ 0.831, p ¼ .181, one-tailed.
¼ .002, two-tailed. With the abstract words,
Downloaded by [201.141.53.213] at 06:48 11 October 2012

H.A. showed the same pattern, with both seman-


tically related arrays (semantically related ¼ 52%; EXPERIMENT 5: ABSTRACTNESS 3
semantically unrelated ¼ 71%), x2(1) ¼ 9.953, p FREQUENCY
¼ .0016, two-tailed, and associatively related
arrays (associatively related ¼ 57%, associatively One possible explanation for the very different
unrelated ¼ 75%), x2(1) ¼ 9.211, p ¼ .002, two- patterns of performance demonstrated by H.A.
tailed. and D.Z. in Experiment 4 could be their overall
Patient D.Z. showed the same pattern with ability to process concrete and abstract words rela-
statistically significant interference effects for tive to A.Z. To explore this possibility, we tested
semantically related concrete words (related ¼ both patient H.A. and patient D.Z. on a large
75%, unrelated ¼ 99%), x2(1) ¼ 32.431, p ¼ number of stimuli that were manipulated by con-
.0001, two-tailed, and associatively related con- creteness and frequency (Experiment 1 in Crutch
crete words (related ¼ 72%, unrelated ¼ 88%), & Warrington, 2005). While previous reports of
x2(1) ¼ 8.308, p ¼ .002, two-tailed. She showed refractory access dysphasia have often reported an
the same pattern for abstract words, whether absence of frequency effects in these patients,
they were semantically related abstract words Crutch and Warrington (2005) demonstrated
(related ¼ 58%, unrelated ¼ 80%), x2(1) ¼ that patient A.Z. showed an unusual reverse fre-
16.253, p ¼ .0001, two-tailed, or associatively quency effect whereby she was actually better
related abstract words (related ¼ 67%, unrelated with spoken-word – written-word matching with
¼ 95%), x2(1) ¼ 24.712, p ¼ .0001, two-tailed. lower frequency words than with matching with
We also tested the possibility that, although high-frequency words. Thus, examining frequency
these patients showed significant differences for effects in patients H.A. and D.Z. might be of
all conditions, the pattern of performance might interest. Moreover, it has been noted (Crutch &
reveal greater impairment on semantically related Warrington, 2010b) that the previous patient
than on associatively related arrays for concrete reported to not replicate A.Z.’s double dissociation
words and greater impairment on associatively with abstract and concrete stimuli did not show a
related than on semantically related arrays for typical concreteness effect (i.e., improved perform-
abstract words. This would be consistent with ance with increasing concreteness, see Hamilton &

2
Given that Crutch and Warrington (2010b) have reported facilitation on reading tasks using abstract and concrete stimuli in
associatively or semantically related arrays, two-tailed tests were used in Experiment 4.

626 COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8)


REFRACTORY ACCESS DYSPHASIA

Coslett, 2008, Experiment 2). Unfortunately, it is Table 1. Patient performance with nouns varying in abstractness
difficult to fully assess concreteness effects in
Frequency
patient UM-103 given the data presented in
Hamilton and Coslett (2008). The words pre- Patient Abstractness LF MF HF Average
sented, which were borrowed from Crutch and
Warrington (2005), were all abstract, varying H.A. VLA 66 63 53 60
LA 59 78 66 68
only in the degree of abstractness. That is, the MA 78 74 78 77
stimuli were not composed of words along the 68 72 66 68
entire continuum of abstract to concrete.
D.Z. VLA 63 72 69 68
Nevertheless, it is important to determine LA 75 88 78 80
whether the patients presented here have a MA 69 94 91 84
similar absence of frequency effects3 (as reported Average 69 84 79 77
with many cases of refractory access dysphasia) or
Note: Crutch and Warrington (2005). VLA ¼ very low
even reversed frequency effects (as in Crutch & abstractness; LA ¼ low abstractness; MA ¼ medium
Warrington’s patient A.Z.) and whether their per-
Downloaded by [201.141.53.213] at 06:48 11 October 2012

abstractness. LF ¼ low frequency; MF ¼ medium frequency;


formance is influenced by degree of abstractness. HF ¼ high frequency.

Method
When examining frequency effects with these
We tested H.A. and D.Z. on the same materials as stimuli, H.A. showed no significant frequency
those used to establish A.Z.’s reverse frequency effect (low frequency ¼ 67.7%, medium fre-
effect (Crutch & Warrington, 2005) to determine quency ¼ 72%, high frequency ¼ 66%), x2(2) ¼
whether H.A. and D.Z. demonstrated an absence 0.9, p ¼ .637, two-tailed. In contrast, D.Z. did
of frequency or concreteness effects in simple show a frequency effect (low frequency ¼ 68.8%,
spoken-word – written-word matching tasks. In medium frequency ¼ 84%, high frequency ¼
this experiment, each word was sampled once in 79.2%), x2(2) ¼ 6.954, p ¼ .03, two-tailed.
each array, with no repetition. However, it is of note that the frequency effect
for D.Z. was atypical, in that she performed non-
significantly better on the medium-frequency than
Results on the high-frequency words, x2(1) ¼ 0.874, p ¼
As can be seen in Table 1, patient H.A. showed .35, two-tailed, and she showed no significant
better performance as words became less abstract, difference between low-frequency and high-fre-
although the effect was only marginal (very low quency words, x2(1) ¼ 2.704, p ¼ .10, two-
abstractness ¼ 60% correct, low abstractness ¼ tailed. The only significant difference was
68%, medium abstractness ¼ 77%), x2(2) ¼ between the low-frequency and the medium-fre-
5.398, p ¼ .067, two-tailed. Patient D.Z.’s per- quency items, x2(1) ¼ 6.531, p ¼ .01, two-tailed.
formance also improved as words became less In summary, H.A. is very similar to other
abstract (very low abstractness ¼ 68% correct, patients with refractory semantic dysphasia reported
low abstractness ¼ 80%, medium abstractness ¼ in the literature, in that he shows an absence of fre-
84%), x2(2) ¼ 8.265, p ¼ .016, two-tailed. quency effects in spoken-word–written-word

3
Many have included the absence of frequency effects among the core features of refractory access dysphasia. However, the pre-
dicted status of frequency effects in refractory access patients is somewhat unclear. Some have argued for weakened or attenuated
frequency effects (in comparison to exaggerated effects of frequency seen in patients with storage deficits) or even an absence of
effects (i.e., no statistically significant effects). However, there have been refractory access patients reported with significant frequency
effects (Warrington & McCarthy, 1983), with an absence of frequency effects (Hamilton & Coslett, 2008) and even reversed fre-
quency effects (Crutch & Warrington, 2005).

COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8) 627


HAMILTON AND MARTIN

matching tasks. Patient D.Z. did show a statistically significantly worse when items were selected
significant frequency effect, but only between low- from arrays composed of semantically related
frequency and medium-frequency words, and not items than when they were from semantically
between low-frequency and high-frequency or unrelated arrays.
between medium-frequency and high-frequency In Experiment 2, both patients showed a
words. This is in contrast to patient A.Z. (Crutch similar pattern of performance when presented
& Warrington, 2005) who showed a reversed fre- with proper nouns related to cities, replicating
quency effect. In addition, H.A. and D.Z. both three previous case studies by Crutch and
showed better performance as words became less Warrington (2003, 2010a) and providing further
abstract using these stimuli, although this trend support for the hypothesis that semantic represen-
failed to reach statistical significance in H.A.’s case. tations related to geography might have a substan-
tial spatial component as proposed by Crutch and
Warrington (2003, 2010a). Of course, it may be
GENERAL DISCUSSION argued that geographically proximate cities prob-
ably share other features in addition to spatially
Downloaded by [201.141.53.213] at 06:48 11 October 2012

Patients with features of refractory access dyspha- defined geography. For example, one might
sia have been important sources of data in making argue that cities sharing geographical proximity
inferences about semantic organization. While (e.g., cities in the north-east US) also share topo-
these studies have motivated several novel hypoth- graphy, fauna, dialect, history, and cuisine that
eses regarding semantic organization, the data might be represented in a nonspatial code.
have come from very few patients. Given that Cleverly designed future work with patients and
the constellation of performance termed “refrac- healthy subjects should be able to arbitrate
tory access dysphasia” may vary across patients, between these two hypotheses.
and some have even questioned its validity (Rapp In Experiment 3, D.Z. and H.A. were tested
& Caramazza, 1993), replication is especially with additional proper nouns, this time with the
important in this area. Thus, we attempted to category of famous people. Again, these patients
replicate a number of findings reported in the lit- were remarkably similar to patients A.Z. (Crutch
erature to determine which patterns are replicable & Warrington, 2004) and UM-103 (Hamilton
in two patients showing the three primary features & Coslett, 2008), in that they showed greater
of refractory access dysphasia. interference when presented with arrays of
Patient D.Z. and H.A. were very similar to famous people drawn from the same occupation
other patients identified as having refractory than when presented with arrays of famous
access dysphasia (patient A.Z. from Crutch & people from different occupations. These data
Warrington, 2003, 2005; patient J.M. from provide further support that occupation may be
Forde & Humphreys, 1995, 1997, 2007) in that one of the primary features along which knowl-
they were both globally aphasic, with very edge of famous people are organized (Crutch &
limited spoken output and similar overall patterns Warrington, 2004).
of impairment in language processing. However, when examining performance on
Experiment 1 revealed both patients to have all abstract and concrete words in Experiment 4,
the hallmarks of refractory semantic access impair- these patients diverged markedly from previously
ment. On spoken-word– written-word matching reported patients A.Z. (Crutch & Warrington,
tasks their performance was highly variable, but 2005) and I.R.Q. (Crutch et al., 2006). Unlike
declined with repetition of the same stimuli. patient A.Z., who yielded no difference between
Performance improved when response – stimulus associatively related concrete words and synony-
intervals increased from 1 s to 30 s. Most impor- mously related abstract words, both patient H.A.
tantly, both patients were extremely sensitive to and patient D.Z. showed statistically significant
semantic relatedness of items, performing differences when concrete and abstract words

628 COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8)


REFRACTORY ACCESS DYSPHASIA

were placed in related arrays, regardless of whether abstract words. As detailed in Hamilton and
the arrays were semantically or associatively Coslett (2008), the subsequent Crutch et al.
related. Thus, patients D.Z. and H.A. demon- (2006) study of patient I.R.Q. was equivocal
strated the same pattern of performance on con- given that close inspection of the data revealed
crete and abstract words as did patient UM-103 variability in performance on unrelated stimuli
previously reported by Hamilton and Coslett that generated the critical null effect for synon-
(2008). ymous abstract words (see Figure 4 of Crutch
Finally, Experiment 5 looked at patient D.Z. et al., 2006). More specifically, the patient was
and H.A.’s performance on stimuli manipulated equally poor on abstract words presented in
for abstractness and frequency. As has been the semantically related arrays as on abstract words
case with many previously reported access patients, presented in associatively related arrays.
H.A. showed no significant effects of word fre- However, the patient’s performance on the unre-
quency. D.Z., on the other hand, did show a stat- lated arrays also differed in these two conditions,
istically significant, although atypical, frequency giving the impression of a null effect for abstract
effect. Specifically, she was significantly better words in semantically related arrays. It is impor-
Downloaded by [201.141.53.213] at 06:48 11 October 2012

with medium-frequency than with low-frequency tant to emphasize that there is no reason to antici-
words, but showed no significant effect between pate any performance difference, other than
low-frequency and high-frequency words or random variation, in the unrelated arrays in these
between medium-frequency and high-frequency two conditions. The effect should be driven by
words. better performance on abstract words organized
In the next section, we evaluate the evidence for in semantically related arrays than on abstract
the “differential frameworks model” and suggest words in associatively related arrays, not in the
hypotheses as to why patients H.A., D.Z., and decreased performance for unrelated arrays.
UM-103 (Hamilton & Coslett, 2008) might More recently, Crutch and Warrington (2010b)
differ from other patients reported by Crutch have offered additional data from patient F.B.I. in
and colleagues. support of a revised version of the differential fra-
meworks hypothesis, which they term the “differ-
ential dependence hypothesis”. F.B.I. is very
Reassessing evidence for the differential
different from the patients presented in this
frameworks model
paper (and patient A.Z.) in that he had sufficient
Crutch and Warrington (2005) first introduced production abilities to allow testing of his
the “differential frameworks model” based on reading. Crutch and Warrington (2010b) conclude
data from patient A.Z. A subsequent study of a that F.B.I. demonstrated the same pattern of
bilingual patient, I.R.Q., was offered as further results as patient A.Z. in spoken-word –written-
support (Crutch et al., 2006). As pointed out in word matching tasks and the analogous pattern
Hamilton & Coslett (2008), the patient data sup- of facilitation in reading tasks. Patient F.B.I. was
porting the differential frameworks model might better at reading concrete words when they were
be considered somewhat equivocal. First, the orig- presented with semantically related words and
inal data rested crucially on a null effect—that is, better at reading abstract words when they were
no differences between related and unrelated presented with associatively related words. In con-
abstract words presented in synonymous arrays trast, he showed no facilitation for associated con-
and no difference between related and unrelated crete words or synonymous abstract words.
concrete words presented in associatively related However, it is important to note that in one of
arrays. It is of note that these null effects were pre- the spoken-word – written-word matching exper-
sented in the context of a very limited number of iments, patient F.B.I. showed only half of the
stimuli—eight sets of four words in each of the crucial dissociation. Specifically, for concrete
related and unrelated conditions for concrete and words, F.B.I. actually showed statistically

COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8) 629


HAMILTON AND MARTIN

significant effects of semantic association pictures. They found that subjects were more
(although subsequent testing with additional likely to fixate pictures that were associated to
stimuli were offered to establish the null effect). auditorily presented abstract words than those
Thus, three patients have been reported to demon- that were associated with auditorily presented con-
strate the pattern supporting a differential frame- crete words. Of note, the same pattern was
works model (to varying degrees), while patients observed when concrete words were presented,
D.Z. and H.A. represent the second and third but to a lesser extent. This suggests that associative
patients that fail to demonstrate the dissociation. relationships were relevant for both concrete and
Other data from the reading errors of patients abstract words, but to a greater extent for abstract
with deep dyslexia have also been reported to words. This pattern of data can be interpreted as
support the differential frameworks hypothesis. supporting a version of the differential dependence
Crutch (2006) performed a post hoc analysis of hypothesis. However, Dunabeitia et al. tested only
semantic reading errors of four deep dyslexic patients associative relatedness and did not evaluate seman-
originally reported by Coltheart, Patterson, and tic similarity or synonymy in the organization of
Marshall (1980). This analysis revealed a greater concrete and abstract words using their method-
Downloaded by [201.141.53.213] at 06:48 11 October 2012

number of associatively related errors than semanti- ology. Also of note, this experiment tested associ-
cally related errors when reading abstract words and ative relationships between an abstract word and a
more semantically related errors for concrete words. concrete concept (the picture), not associative
However, it is of note that the Crutch (2006) analysis relationships between two abstract concepts.
judged over 35% of the reading errors to abstract Crutch and Jackson (2011) recently examined
words to be semantically similar rather than associ- the role of semantic similarity and association
ated. Of further note, poorer performance on with concrete and abstract words using an odd-
abstract words than on concrete words is a near-uni- one-out paradigm with normal subjects. Subjects
versal feature of deep dyslexia. Indeed, the four were presented four words in a single display and
patients reported by Crutch (2006) made nearly selected which item did not fit with the other
three times as many errors on abstract words than items. Abstract, middle-concreteness, and con-
on concrete words. Given deep dyslexics’ particular crete stimuli were selected and placed in related
difficulty with abstract words, it seems likely that a arrays. These arrays were rated for both semantic
large percentage of errors for these patients might association and semantic similarity. Using a mul-
merely be concrete responses to abstract targets, tiple regression approach, the authors found a stat-
which are necessarily associated. Inspection of the istically significant interaction between semantic
Coltheart et al. (1980) data reveals many such similarity and concreteness such that reaction
errors (e.g., phase  moon; amount  money; times (RTs) were faster for concrete stimuli as
condemn  prison).Whether such errors reflect semantic similarity increased. For associative relat-
the organization of abstract semantics or simply a edness, RTs decreased as words became more
bias toward concrete words in deep dyslexia is abstract, although the critical interaction was not
unclear. statistically significant (p ¼ .07). Crutch and
Finally, it is of note that behavioural data from Jackson (2011) also adapted the stimuli for use
healthy subjects has recently been marshalled as with a spoken-word – written-word matching
further evidence for the differential frameworks task administered to N.B.C., a patient with
model (Crutch, Connell, & Warrington, 2009; semantic refractory access dysphasia. This patient’s
Crutch & Jackson, 2011;Dunabeitia, Aviles, pattern of performance on the spoken-word –
Afonson, Scheepers, & Carreiras, 2009). written-word matching task was similar to the
Dunabeitia et al. (2009) measured eye movements pattern reported for normal subjects in the odd-
in a visual-world paradigm. They auditorily pre- one-out tasks. However, inspection of patient
sented abstract or concrete Spanish words that N.B.C.’s data reveals the same problem as that
were associated with a target in an array of four outlined above for patient I.R.Q. (Crutch et al.,

630 COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8)


REFRACTORY ACCESS DYSPHASIA

2006). One of the critical null effects was driven by of associative and feature-based semantic
worse performance in an unrelated condition— priming). This literature would seem to be
whether abstract words were presented in low- especially relevant to the original Crutch and
similarity or high-similarity arrays made no differ- Warrington (2005) account, given that priming in
ence to the patient’s performance. The null effect normal subjects has been explained in terms of
is the result of random variability in performance “spreading activation” models (Collins & Loftus,
on the unrelated condition (see Crutch & 1975), just as Crutch and Warrington (2005) used
Jackson, 2011, Figure 3). spreading activation to account for refractory
In addition, Crutch and Jackson (2011) stated: effects. Moreover, studies have reported “mediated
“The observation of patients showing association priming”, such that lion primes stripes, presumably
and similarity effects for both abstract and concrete by way of tiger, which would appear to be necess-
words is not capable of refuting or negating this arily mediated by associative connections among
evidence from healthy individuals.” However, we concrete words (Chwilla & Kolk, 2002).
would argue that this depends on the strength of Moreover, Shelton and Martin (1992) have
such evidence from healthy individuals. Failing argued that much of automatic semantic priming
Downloaded by [201.141.53.213] at 06:48 11 October 2012

to replicate the original pattern with patients (all is actually associative priming. Similarly, the
of which show remarkable overlap with the other review of the literature by Hutchison (2003) came
patients reported in the literature other than to the conclusion that automatic semantic priming
their performance on concrete and abstract does not occur for category coordinates but does
stimuli) invites scrutiny of the behavioural occur for associates and for some semantic relations
studies with normal subjects. Our examination of like synonymy. Thus, there also exist data from
these studies reveals questions concerning the normal subjects suggesting that concrete words are
interpretation of some of those data. organized by both semantic similarity and associat-
Finally, returning to the data presented here, we ive relationships. Whether the relative influence of
also found no evidence for a version of the differen- semantic and associative relationships varies across
tial dependence hypothesis that predicts only a rela- the entire continuum of abstract to concrete
tive difference between concrete and abstract words remains to be demonstrated in a priming paradigm.
and their reliance on associative or similarity-based
relationships. For example, if concrete words
What might account for the differences
depend more strongly on similarity (but also have
across patients?
weaker associative connections), and abstract
words depend more strongly on association (but Although the patients reported here did not show
also have weaker connections based on similarity), the pattern expected by the differential frame-
one might anticipate significant differences works hypothesis, one might question whether
between concrete and abstract words in both asso- there is some way to accommodate the present
ciatively related and semantically related arrays, findings that would be in accord with that hypoth-
but larger effects for semantically related concrete esis. Of course, finding a different pattern of per-
arrays than for associatively related concrete arrays formance for some patients does not necessarily
and stronger effects for associatively related abstract undermine the theoretical conclusions drawn
arrays than for semantically related abstract arrays. from contrasting patterns shown by other patients.
However, neither of the patients in this study For example, finding that some patients show a
revealed any statistically significant differences pattern of phonological dyslexia does not under-
between any of the related conditions. mine the conclusion of a role for sublexical
It is of interest that there exists a substantial lit- letter – sound conversion deduced from the
erature reporting priming for associatively related reading pattern shown by surface dyslexics.
words, including concrete words (e.g., bread However, both sets of data need to be accommo-
primes butter; see Hutchison, 2003, for a review dated by the theoretical framework. Currently, it

COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8) 631


HAMILTON AND MARTIN

is unclear how the differential frameworks could in patients D.Z. and H.A. might obscure differ-
account for the equivalent effects for semantically ences that were more apparent with A.Z.’s
and associatively related terms for abstract and weaker refractory effects. Similarly, this account
concrete words that were reported here. Below might explain the effects of associative and seman-
we consider some possible ways in which the tic relatedness observed in D.Z. and H.A. with
current findings might be accommodated. abstract words. We presently have no data that
Conceding for a moment the assumptions of the speak directly to this possibility.
differential frameworks hypothesis, one reason However, given that patient A.Z. (Crutch &
these patients might differ is variability between Warrington, 2005), patient UM-103 (Hamilton &
patients within the core features of “refractory Coslett, 2008), and patients D.Z. and H.A. have
semantic access dysphasia”. On one hand, given all been tested with the same concrete and abstract
that such patients have large lesions, encompassing stimuli from Crutch and Warrington (2005;
frontal, parietal, and temporal lobes, and are glob- Experiment 4 here), we can rule out the possibility
ally aphasic, the correspondence across patients that differences in overall severity of impairment
with regard to the three core features outlined are responsible for the disparate patterns. While
Downloaded by [201.141.53.213] at 06:48 11 October 2012

above is rather remarkable. Moreover, many of patient UM-103 (Hamilton & Coslett, 2008) was
these patients show very similar patterns of per- more severely impaired on both concrete and
formance across multiple semantic domains (i.e., abstract words than was A.Z., patient H.A.’s
geography, famous people, etc.). However, it is overall performance appears to be very similar to
not surprising that these patients might differ on that of A.Z., as does D.Z.’s (whose performance is
dimensions such as degree of impairment on numerically better than that of A.Z.).
abstract concepts or the presence (patient D.Z.), Finally, it is of note that patient A.Z. (Crutch
absence (patient H.A.), or reversal of frequency & Warrington, 2005), the patient providing the
effects (patient A.Z.; Crutch & Warrington, best patient evidence for the differential frame-
2005). It is possible that variability on these and works account, is unique in that she actually
other dimensions might reflect qualitative differ- showed a reverse frequency effect—she was sig-
ences in access disorders among patients. nificantly more accurate with low-frequency
More specifically, one possibility is that the words. While weak or complete absence of fre-
temporal dynamics of refractory states might quency effects is typical with cases of refractory
differ across patients, such that it may take access dysphasia, A.Z. represents the only reported
longer for some patients to emerge from a refrac- case of reverse frequency effects. Thus, one might
tory state. For example, in the current experiments, posit that there exist other unique features of
we used a 30-s RSI (as used in Forde & A.Z.’s deficit that result in the original dissociation
Humphreys, 1995, and Warrington & between concrete and abstract words.
McCarthy, 1983) to demonstrate an effect of pres- One other possible means of accounting for the
entation rate after finding that a 10-s RSI (as differences between the patients could be to
employed by Crutch & Warrington, 2005) was assume differences in the functional locus of the
not sufficient to reveal differences. Thus, one deficit in terms of the level of word representation
might speculate that the patients reported here that is impaired. In the priming literature, it has
have a longer or more dense refractory state that long been suggested that associative and semantic
allows spreading activation to reach represen- priming occur at different levels of representation.
tations that are more distantly organized by associ- While semantic priming is thought to occur at the
ative or semantic relatedness for both concrete and level of semantic representation, associative
abstract concepts. If we are to assume that concrete priming has been argued to be based in co-occur-
words are most closely organized by semantic rence at the lexical level (Moss, Hare, Day, &
relatedness but also have more distant associative Tyler, 1994; Shelton & Martin, 1992). However,
connections, the more persistent refractory states it has often been assumed that effects of semantic

632 COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8)


REFRACTORY ACCESS DYSPHASIA

relatedness observed in patients with refractory demonstrating associative priming for concrete
semantic access dysphasia must occur at a semantic words) suggest that both concrete and abstract
level, although most of these studies (including the words are linked both via associative connections
present experiments) have not reported data from and by available semantic features. Although
multiple modalities (i.e., pictures and words) or patients showing a different pattern of deficits
multiple tasks. Thus, that these effects occur at from that in previously reported cases do not
some other level of representation (i.e., lexical) necessarily undermine a particular theoretical
has rarely been ruled out. The exception is approach, it is nonetheless necessary for this theor-
patient J.M. (Ferrand & Humphreys, 1996; etical approach to find some means of accommo-
Forde & Humphreys, 1995, 1997, 2007). J.M. dating the disparate findings. At present, it is
was tested extensively on variations of spoken- unclear to us how the similar effects of semantic
word matching tasks and showed interference and associative relations for both concrete and
effects that spread from words to pictures (Forde abstract terms could be accommodated by the
& Humphreys, 1997) and across languages (J.M. differential frameworks hypothesis, as currently
was fluent in six languages), providing strong evi- specified. Above we have made some suggestions
Downloaded by [201.141.53.213] at 06:48 11 October 2012

dence of a semantic locus of the effects (Ferrand & regarding how the differing results across different
Humphreys, 1996). It is possible that refractory patient might be accommodated. Whether or not
access dysphasia patients differ in damage to separ- these possibilities are correct and whether they
able functional foci (i.e., lexical vs. semantic) and would be consistent with the differential frame-
that these deficits might interact with overall works hypothesis can only be determined
damage to semantic representations for abstract through further investigation.
and concrete words, yielding different patterns of
performance on associatively and semantically Manuscript received 2 December 2010
related concrete and abstract words. Whether or Revised manuscript received 5 July 2011
Revised manuscript accepted 18 July 2011
not identifying the functional locus of
First published online 10 November 2011
damage would provide evidence supporting the
differential frameworks hypothesis or a hypothesis
of a common framework with equivalent REFERENCES
weight for associative and semantic relations for
Chwilla, D. J., & Kolk, H. H. J. (2002). Three-step
abstract and concrete words cannot be determined
priming in lexical decision. Memory & Cognition,
at this point. It seems clear, however, that future
30, 217–225.
work in this area must attempt to more clearly Cipolotti, L., McNeil, J., & Warrington, E. (1993).
define the locus of these refractory effects, thus Spared written naming of proper nouns: A case
facilitating comparison of patients across studies. report. Memory, 4, 289–311.
Cipolotti, L., & Warrington, E. K. (1995). Towards a
unitary account of access dysphasia: A single case
CONCLUSIONS study. Memory, 3, 309–332.
Collins, A. C., & Loftus, E. F. (1975). A spreading-
In summary, it would appear that the evidence for activation theory of semantic processing.
the “differential frameworks” hypothesis is pre- Psychological Review, 82, 407–428.
Coltheart, M., Patterson, K., & Marshall, J. C. (1980).
sently equivocal and should be evaluated more
Deep dyslexia. London, UK: Routledge and Kegan Paul.
closely. The patients reported here provide no evi- Crutch, S. J. (2006). Qualitatively different semantic
dence that abstract words are represented primarily representations for abstract and concrete words:
by associative relationships, whilst concrete words Further evidence from semantic reading errors of
are represented primarily by taxonomic similarity deep dyslexic patients. Neurocase, 12(2), 91–97.
(Crutch et al., 2009). Instead, the present data Crutch, S. J., Connell, S., & Warrington, E. K.
(as well as a substantial priming literature (2009). The different representational frameworks

COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8) 633


HAMILTON AND MARTIN

underpinning abstract and concrete knowledge: Implications for access-storage distinctions and the
Evidence from odd-one-out judgements. Quarterly nature of semantic memory. Cognitive
Journal of Experimental Psychology, 62, 1377–1390. Neuropsychology, 14, 367–402.
Crutch, S. J., & Jackson, E. C. (2011). Contrasting Forde, E. M. E., & Humphreys, G. W. (2007).
graded effects of semantic similarity and association Contrasting effects of repetition across tasks:
across the concreteness spectrum. The Quarterly Implications for understanding the nature of refrac-
Journal of Experimental Psychology, 64(7), 1388–1408. tory behavior and models of semantic memory.
Crutch, S. J., Ridha, B. H., & Warrington, E. K. (2006). Cognitive, Affective & Behavioral Neuroscience, 7,
The different frameworks underlying abstract and 198–211.
concrete knowledge: Evidence from a bilingual Hamilton, A. C., & Coslett, H. B. (2008). Refractory
patient with a semantic refractory access dysphasia. access disorders and the organization of concrete
Neurocase, 12, 151–163. and abstract semantics: Do they differ? Neurocase,
Crutch, S. J., & Warrington, E. K. (2003). The selective 14, 131–140.
impairment of fruit and vegetable knowledge: A mul- Hodges, J. R., Patterson, K., Oxbury, S., & Funnell, E.
tiple processing channels account of fine-grain category (1992). Semantic dementia. Progressive fluent
specificity. Cognitive Neuropsychology, 20, 355–372. aphasia, with temporal lobe atrophy. Brain, 115,
Downloaded by [201.141.53.213] at 06:48 11 October 2012

Crutch, S. J., & Warrington, E. K. (2004). The semantic 1783–1806.


organization of proper nouns: The case of people and Howard, D., & Patterson, K. (1992). The Pyramids and
brand names. Neuropsychologia, 42, 584–596. Palm Trees Test. Bury St. Edmunds, UK: Thames
Crutch, S. J., & Warrington, E. K (2005). Abstract and Valley Test Company.
concrete concepts have structurally different rep- Hutchison, K. A. (2003). Is semantic priming due to
resentational frameworks. Brain, 128, 615–627. association strength or feature overlap? A micro-
Crutch, S. J., & Warrington, E. K. (2010a). Spatially- analytic review. Psychonomic Bulletin & Review, 10,
coded semantic information about geographical 785–813.
terms. Neuropsychologia, 48, 2120–2129. James, C. T. (1975). The role of semantic information in
Crutch, S. J., & Warrington, E. K. (2010b). The lexical decisions. Journal of Experimental Psychology:
differential dependence of abstract and concrete Human Perception and Performance, 104, 130–136.
words upon associative and similarity-based infor- Jefferies, E., Baker, S. S., Doran, M., & Lambon, M. A.
mation: Complementary semantic interference and (2007). Refractory effects in stroke aphasia: A conse-
facilitation effects. Cognitive Neuropsychology, 27, quence of poor semantic control. Neuropsychologia,
46–71. 45, 1065–1079.
De Haan, E. H. F., Young, A., & Newcombe, F. (1987). Kay, J., Lesser, R., & Coltheart, M. (1992).
Faces interfere with name classification in a proso- Psycholinguistic Assessments of Language Processing in
pagnosic patient. Cortex, 23, 309–316. Aphasia (PALPA). London, UK: Lawrence
Dunabeitia, J. A., Aviles, A., Afonson, O., Scheepers, Erlbaum Associates.
C., & Carreiras, M. (2009). Qualitative differences Moss, H. E., Hare, M. L., Day, P., & Tyler, L. K.
in the representation of abstract versus concrete (1994). A distributed memory model of the associat-
words: Evidence from the visual-world paradigm. ive boost in semantic priming. Connection Science, 6,
Cognition, 110, 284–292. 413–427.
Dunn, L. M., & Dunn, L. (1981). Peabody Picture Paivio, A. (1986). Mental representations: A dual coding
Vocabulary Test– Revised manual for Forms L and approach. London, UK: Oxford University Press.
M. Circle Pines, MN: American Guidance Service. Rapp, B., & Caramazza, A. (1993). On the distinction
Ferrand, L., & Humphreys, G. (1996). Transfer of between deficits of access deficits of storage: A ques-
refractory states across languages in a global aphasic tion of theory. Cognitive Neuropsychology, 10, 113–141.
patient. Cognitive Neuropsychology, 13, 1163–1191. Rorden, C., Karnath, H.-O., & Bonilha, L. (2007).
Forde, E. M. E., & Humphreys, G. W. (1995). Improving lesion –symptom mapping. Journal of
Refractory semantics in global aphasia: On semantic Cognitive Neuroscience, 19, 1081–1088.
organization and the access-storage distinction in Semenza, C. (2009). The neuropsychology of proper
neuropsychology. Memory, 3, 265–307. names. Mind & Language, 24, 347–369.
Forde, E. M. E., & Humphreys, G. W. (1997). A Shelton, J. R., & Martin, R. C. (1992). How semantic is
semantic locus for refractory behaviour: automatic semantic priming? Journal of Experimental

634 COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8)


REFRACTORY ACCESS DYSPHASIA

Psychology: Learning, Memory & Cognition, 18, Journal of Experimental Psychology: Learning,
1191–1210. Memory & Cognition, 25, 1256–1271.
Shewan, C. M., & Kertesz, A. (1980). Reliability and Warrington, E. K., & Cipolotti, L. (1996). Word com-
validity characteristics of the Western Aphasia prehension: The distinction between refractory and
Battery (WAB). Journal of Speech and Hearing storage impairments. Brain, 119, 611–625.
Disorders, 45(3), 308–324. Warrington, E. K., & McCarthy, R. (1983). Category
Walker, I., & Hulme, C. (1999). Concrete words are specific access dysphasia. Brain, 106, 859–878.
easier to recall than abstract words: Evidence for Warrington, E. K., & Shallice, T. (1984). Category
semantic contribution to short-term serial recall. specific semantic impairments. Brain, 107, 829–854.

APPENDIX

Stimuli used in Experiment 1


Downloaded by [201.141.53.213] at 06:48 11 October 2012

Columns were used to construct related arrays; rows were used


to construct unrelated arrays.

Stimuli used in Experiment 1

Reptiles/
Fruits Vehicles Amphibians Instruments
orange truck snake guitar
apple car turtle drum
lemon boat frog trumpet
strawberry train lizard flute

COGNITIVE NEUROPSYCHOLOGY, 2010, 27 (8) 635

You might also like